On-line signature verification with hidden Markov models

This paper addresses the problem of online signature verification based on hidden Markov models (HMM). We use a novel type of digitizer tablet and pay special attention to the use of pen-tilt. We investigate the verification reliability based on different forgery types. We compare the discriminative value of the different features based on a linear discriminant analysis (LDA) and show that pen-tilt is important. On the basis of home-improved, over-the-shoulder and professional forgeries, we show that the amount of dynamic information available to an imposter is important and that forgeries based on paper copies are easier to detect. The results obtained with a database of almost 5000 signatures of 51 persons with highly skilled forgeries include equal-error rates between 1% and 1.9%.